We Built a GitHub Trending Page That Actually Uses Data
GitHub's trending page is broken. So we built our own — powered by 10.5 billion GitHub events, with filters, language breakdowns, and real activity metrics.
GitHub's trending page has a problem: nobody knows how it works.
The ranking algorithm is opaque. The time windows are vague. Projects appear and disappear without explanation. And if you've ever wondered why a repo with 200 stars shows up above one with 50,000 — well, so has everyone else.
We decided to build something different.
What We Shipped
ossinsight.io/trending — a trending page built on top of 10.5 billion GitHub events, with transparent ranking and real filters.
Here's what you get:
- Period filters — Today, This Week, This Month. Pick your window, see different results.
- Language filters — Filter by Python, TypeScript, Rust, Go, or any language. See what's trending in your ecosystem, not everyone else's.
- Real metrics — Stars, forks, and the actual activity numbers. No mystery algorithm.
- Deep links — Every repo links to its full OSSInsight analysis page with contributor graphs, commit history, and community health metrics.
What's Trending Right Now (March 2026)
Here's what the page shows today:
| Repo | Stars | Language | What It Does |
|---|---|---|---|
| obra/superpowers | 109K | Shell | Agentic skills framework for software development |
| affaan-m/everything-claude-code | 103K | JavaScript | Agent harness optimization — skills, memory, security |
| anthropics/skills | 101K | Python | Official Anthropic agent skills repository |
| karpathy/autoresearch | 53K | Python | AI agents running research on single-GPU training |
| bytedance/deer-flow | 41K | Python | Open-source SuperAgent — researches, codes, creates |
| gsd-build/get-shit-done | 40K | JavaScript | Meta-prompting and spec-driven development |
| TauricResearch/TradingAgents | 40K | Python | Multi-agent LLM financial trading framework |
| shareAI-lab/learn-claude-code | 38K | TypeScript | Build a nano Claude Code from scratch |
The pattern is hard to miss: 7 out of the top 8 trending repos are AI agent projects. Skills frameworks, agent harnesses, autonomous research tools. This is what developers are building and starring right now.
Why Not Just Use GitHub Trending?
Three reasons:
1. Transparency. GitHub's trending algorithm is a black box. Ours is simple: we rank by recent star velocity and activity events. You can see the numbers.
2. Filters that work. GitHub trending lets you filter by language and time period, but the results often feel arbitrary. Our filters are deterministic — same inputs, same outputs.
3. Context. A repo name and star count tells you almost nothing. Every entry on our trending page links to a full analysis: contributor breakdown, commit velocity, issue health, fork patterns. You can go from "what's popular" to "is this actually healthy" in one click.
The Data Behind It
The trending page draws from the same dataset that powers all of OSSInsight: 10.5 billion GitHub events collected in real time. Every push, star, fork, issue, and PR across all of public GitHub.
When you filter by "This Week" and "Python," you're not seeing a curated list. You're seeing what the data says — which Python repos had the most activity in the last 7 days, measured by real events.
What's Next
This is v1. We're planning:
- Trending developers — not just repos, but who's shipping the most interesting work
- Category trends — trending within collections (AI, databases, DevOps, etc.)
- Historical snapshots — what was trending last month, last quarter, last year
- RSS feeds — subscribe to trending in your language
Try It
Filter by your language. Click into a repo that catches your eye. Go deeper than stars.
OSSInsight analyzes 10.5B+ GitHub events in real time. Explore trending repos, collections, and developer analytics at ossinsight.io.